5
$\begingroup$

If $\alpha,\beta,\gamma \in [-3,10].$ Then largest value of the determinant

$$\begin{vmatrix}3\alpha^2&\beta^2+\alpha\beta+\alpha^2&\gamma^2+\alpha\gamma+\alpha^2\\\\ \alpha^2+\alpha\beta+\beta^2& 3\beta^2&\gamma^2+\beta\gamma+\beta^2\\\\ \alpha^2+\alpha\gamma+\gamma^2& \beta^2+\beta\gamma+\gamma^2&3\gamma^2\end{vmatrix}$$

Try: I am trying to break that determinant into product of 2 determinants but not able to break it.

Could someone help me in this question? Thanks.

$\endgroup$
7
+50
$\begingroup$

The matrix is the product $AB$, where $$A=\begin{pmatrix}\alpha^2&\alpha&1\\\beta^2&\beta&1\\ \gamma^2&\gamma&1\end{pmatrix}\mbox{ and }B=\begin{pmatrix}1&1&1\\\alpha&\beta&\gamma\\ \alpha^2&\beta^2&\gamma^2\end{pmatrix}.$$ So its determinant is the product of $\det(A)$ and $\det(B)$. The matrix $A$ can be obtained from $B$ by exchanging the first and the third row and then transposing the result. Therefore $\det(A)=-\det(B)$ and the determinant of the given matrix is $-\det(B)^2$. This is a non-positive expression and its maximal value $0$ is attained if and only if $\det(B)=0$.

Now $B$ is a Vandermonde matrix and its determinant is known to be $(\gamma-\beta)(\gamma-\alpha)(\beta-\alpha)$. This determinant is zero and thus the given determinant attains its maximal value if and only if two or more of the numbers $\alpha,\beta,\gamma$ are equal.

$\endgroup$
  • $\begingroup$ Beautiful solution, but how did you obtain $A$ and $B$? Was it just inspiration, or previous experience? $\endgroup$ – Alex M. May 3 '19 at 10:15
  • 2
    $\begingroup$ To be honest, a) I had seen the value of the determinant in the other answers and it is related to Vandermonde, b) the question speaks of breaking into a product, c) the squares in the matrix hint to Vandermonde and d) the fact that every entry is a sum of three terms (sometimes equal) also hints a product of Vandermonde-like matrices. The rest is experimenting... $\endgroup$ – Helmut May 3 '19 at 12:34
4
$\begingroup$

Hint: Using the rules of SARRUS we get after simplifying $$- \left( \alpha-\gamma \right) ^{2} \left( \beta-\gamma \right) ^{2} \left( \beta-\alpha \right) ^{2} $$

$\endgroup$
  • $\begingroup$ I think so! What result do you get? $\endgroup$ – Dr. Sonnhard Graubner Apr 27 '19 at 15:22
  • $\begingroup$ This is nice to hear! $\endgroup$ – Dr. Sonnhard Graubner Apr 27 '19 at 15:23
  • $\begingroup$ A big typo is only a typo, big or not! $\endgroup$ – Dr. Sonnhard Graubner Apr 27 '19 at 15:24
  • $\begingroup$ Answer given is $0$ $\endgroup$ – DXT Apr 28 '19 at 1:19
  • $\begingroup$ I would eat a jar of fish oil rather than performing all those multiplications, and then all the simplifications. Plus, you are cheating, because you haven't performed them yourself, but rather used a computer algebra system to do them. $\endgroup$ – Alex M. Apr 30 '19 at 14:18
4
$\begingroup$

Being a $3 \times 3$ determinant, it is too small to be attacked with sophisticated techniques, but also too involved to be attacked by brute force. We shall compute it with successive simplifications of its rows and columns. In the following, $R_i$ and $C_j$ will mean the $i$-th row and the $j$-th column, and $[R_i \to aR_i + bR_j]$ means to replace the $i$-th row with the linear combination $aR_i + bR_j$ where $a$ and $b$ are numbers. The computation, then, goes as follows:

$$\begin{align*} & \begin{vmatrix} 3\alpha^2 & \beta^2 + \alpha \beta + \alpha^2 & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha^2 + \alpha \beta + \beta^2 & 3\beta^2 & \gamma^2 + \beta \gamma + \beta^2 \\ \alpha^2 + \alpha \gamma + \gamma^2 & \beta^2 + \beta \gamma + \gamma^2 & 3\gamma^2\end{vmatrix} = [R_2 \to R_2 - R_3] = \\[10pt] & \begin{vmatrix} 3\alpha^2 & \beta^2 + \alpha \beta + \alpha^2 & \gamma^2 + \alpha \gamma + \alpha^2 \\ (\beta - \gamma) (\alpha + \beta + \gamma) & (\beta - \gamma) (2\beta + \gamma) & (\beta - \gamma) (\beta + 2\gamma) \\ \alpha^2 + \alpha \gamma + \gamma^2 & \beta^2 + \beta \gamma + \gamma^2 & 3\gamma^2\end{vmatrix} = \\[10pt] (\beta - \gamma) & \begin{vmatrix} 3\alpha^2 & \beta^2 + \alpha \beta + \alpha^2 & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha + \beta + \gamma & 2\beta + \gamma & \beta + 2\gamma \\ \alpha^2 + \alpha \gamma + \gamma^2 & \beta^2 + \beta \gamma + \gamma^2 & 3\gamma^2\end{vmatrix} = [R_3 \to R_3 - R_1] = \\[10pt] (\beta - \gamma) & \begin{vmatrix} 3\alpha^2 & \beta^2 + \alpha \beta + \alpha^2 & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha + \beta + \gamma & 2\beta + \gamma & \beta + 2\gamma \\ (\gamma - \alpha) (\gamma + 2\alpha) & (\gamma - \alpha) (\alpha + \beta + \gamma) & (\gamma - \alpha) (2\gamma + \alpha) \end{vmatrix} = \\[10pt] (\beta - \gamma) (\gamma - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta^2 + \alpha \beta + \alpha^2 & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha + \beta + \gamma & 2\beta + \gamma & \beta + 2\gamma \\ \gamma + 2\alpha & \alpha + \beta + \gamma & 2\gamma + \alpha \end{vmatrix} = [C_2 \to C_2 - C_1] = \\[10pt] (\beta - \gamma) (\gamma - \alpha) & \begin{vmatrix} 3\alpha^2 & (\beta - \alpha) (\beta + 2\alpha) & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha + \beta + \gamma & \beta - \alpha & \beta + 2\gamma \\ \gamma + 2\alpha & \beta - \alpha & 2\gamma + \alpha \end{vmatrix} = \\[10pt] (\beta - \gamma) (\gamma - \alpha) (\beta - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta + 2\alpha & \gamma^2 + \alpha \gamma + \alpha^2 \\ \alpha + \beta + \gamma & 1 & \beta + 2\gamma \\ \gamma + 2\alpha & 1 & 2\gamma + \alpha \end{vmatrix} = [C_3 \to C_3 - C_1] \\[10pt] (\beta - \gamma) (\gamma - \alpha) (\beta - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta + 2\alpha & (\gamma - \alpha) (\gamma + 2\alpha) \\ \alpha + \beta + \gamma & 1 & \gamma - \alpha \\ \gamma + 2\alpha & 1 & \gamma - \alpha \end{vmatrix} = \\[10pt] (\beta - \gamma) (\gamma - \alpha)^2 (\beta - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta + 2\alpha & \gamma + 2\alpha \\ \alpha + \beta + \gamma & 1 & 1 \\ \gamma + 2\alpha & 1 & 1 \end{vmatrix} = [C_3 \to C_3 - C_2] \\[10pt] (\beta - \gamma) (\gamma - \alpha)^2 (\beta - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta + 2\alpha & \gamma - \beta \\ \alpha + \beta + \gamma & 1 & 0 \\ \gamma + 2\alpha & 1 & 0 \end{vmatrix} = \\[10pt] - (\beta - \gamma)^2 (\gamma - \alpha)^2 (\beta - \alpha) & \begin{vmatrix} 3\alpha^2 & \beta + 2\alpha & 1 \\ \alpha + \beta + \gamma & 1 & 0 \\ \gamma + 2\alpha & 1 & 0 \end{vmatrix} = \\[10pt] - (\beta - \gamma)^2 (\gamma - \alpha)^2 (\beta - \alpha)^2 \end{align*} $$

where the last $3 \times 3$ determinant has been expanded along the $3$rd column.

Since the determinant is the negative of the square of a product of real numbers, its maximum will be $0$, reached whenever any two of $\alpha, \beta, \gamma \in [-3,10]$ are equal.

$\endgroup$
2
$\begingroup$

Hint: The matrix is symmetric. You can use the Cholesky decomposition to rewrite the matrix $A$ as $A=LL^T$, in which $L$ is a lower triangular matrix. Then the determinant is $\det A = \det L \det L^T = \det^2 L$. Then $\det L$ is given by the product of the diagonal elements of $L$ as it is a lower triangular matrix.

$\endgroup$
  • $\begingroup$ Yeah, well, saying it is easy, but why don't you also do it? Finding $L$ doesn't seem to be easier that computing the determinant by brute force, so why bother, then? $\endgroup$ – Alex M. Apr 30 '19 at 14:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.